Artificial Intelligence and Search Engines. Part One: Changing the Way We Search

Over the course of our two-part blog series, we will take an in-depth look at the growth and progression of artificial intelligence in relation to search technology. In Part One we will examine how artificial intelligence is changing the way that people utilise search engines. In Part Two we take a look at what these changes mean for the SEO industry and the way your website ranks on Google.

Artificial intelligence (AI) has been a buzzword around SEO for some time now. For a number of years, SEO professionals have postulated that AI would have a huge role to play in the future of the industry. But this all looked a long way off in 2012, when Google was struggling to get its most advanced AI software to even recognise whether or not a cat was present in a video.

Over the past few years there have marked improvements in the technology to the point where we are seeing AI systems integrate seamlessly into people’s lives. AI assistants on mobile devices and computers have become commonplace – Amazon’s Alexa and Apple’s Siri are just two examples attempting to show how clever they are. AI has slowly been colonising search engines too. And while it’s certainly no secret, the growing influence of AI on Google has taken some search professionals by surprise. But to understand just how AI is affecting the landscape of SEO, we must first establish the ways that AI is influencing and altering the ways that people use search engines every day.

Understanding artificial intelligence

AI may well be the future of search, and therefore being able to understand AI and interpret how it works is an important challenge for businesses and search engine professionals. Firstly, it is useful to understand what AI actually is and how it is applied to search engines.

When we talk about AI, we are referring to the attempts to make machines do the things that human minds can do. Machines are typically extremely efficient at being able to do the things that they are told to do, but cannot do things where they are asked to make decisions or utilise skills such as creativity or problem-solving, which we think of as intrinsically human.

Essentially it is asking machines to learn from their own experiences, rather than learning from things that they have been told by their programming. This has taken many forms, from computers that have learned to play chess better than Grandmasters to cars with the ability to drive themselves.

AI is used within search engines a number of ways. The most of obvious use is in the augmenting of algorithms used to rank websites in order to more effectively offer high quality search results. But AI is also used in a variety of situations that you might not realise when you browse on your phone or search on your laptop.
To understand these, we need to take a closer look at the different types of artificial intelligence that are used by Google, Microsoft, Apple and other technology giants who are leading the field.

The progression of artificial intelligence

There is a lot of talk of AI in modern settings, but when you attempt to understand how it is used within search engines you come up against the problem that the term ‘artificial intelligence’ is simply far too broad. AI can simultaneously refer to the earliest beginnings of the concept, the modern day technology that influences our lives right now, and the myriad possibilities for the future. It is important, then, to break down the phrase into more manageable concepts.

You might not realise that AI dates back to 1956 when computer scientists coined the phrase. In its infancy, AI pioneers dreamed primarily of machines that could mimic all aspects of human intellect – think of androids in science fiction films. Today, however, those at the forefront of the technology realise that the advantages of AI actually come from much narrower forms.

Machine learning

Rather than trying to produce AI that mimics the human experience, pioneers look for opportunities where AI can perform difficult or time-consuming tasks that we previously have relied on human minds for.

This is where we first encounter the phrase ‘machine learning’. Machine learning is a form of AI where a computer is programmed with algorithms to learn and improve at task by being shown a large amount of data. For example, if you could program a computer to understand what letters look like, it could then scan images for words. While impressive feats were accomplished with this form of machine learning, it would still encounter problems.

For example, if the first letter of a word was partially obscured, leaving the letters “E-L-L-O” a computer would not be able to recognise what most English-speaking humans would know instantly: the word is ‘hello’. It was for the same reason that Google was struggling to get its AI to recognise a cat in a video – it worked some of the time, but if the images were obscured, the AI simply was not advanced enough to recognise cats effectively.

Deep learning

It was relatively recently that the concept of ‘deep learning’ was introduced. This is a form of machine learning in which computers are exposed to a huge number of images and use something called neural networks, which examine different elements of the image to ascertain what it is. With the advent and advancement of modern computers, it is now possible for a computer to analyse millions of images and learn from its mistakes. Deep learning is currently the most effective form of AI.

This shows us, then, that what we truly need for successful AI is actually a very large amount of data that can be fed into a machine and allow it to learn. This is, in part, why Google is able (and will continue to be able) to use AI so effectively – the company simply has access to such a huge amount of extremely rich data to provide to its AI systems.

Revolutionising the way we search

So, what does the use of deep learning AI mean for the future of search? The truth is it already having an enormous impact in ways that you might not even have noticed. For example, if you have ever searched using Google Assistant (starting by saying the phrase ‘OK Google’ aloud) you may have watched your device interpreting the words you are saying. It is cleverly able to ignore hesitations and recognise natural speech patterns.

For a glimpse into just how advanced Google’s speech recognition has become, take a look at this video where Google Assistant takes a query from a user about making a women’s hair appointment and then actually makes a call to the salon in order to get the appointment booked, mimicking human speech and understanding the nuances of the conversation.

Interestingly, Google has even able to understand the difference between the ways that different generations interact with it. For example, Generation X learned to use search engines in a way that was functional, for example typing in a query like “cheap Spain holidays”. On the other hand, Millennials now interact with search engines with more natural questions like “where is cheap to go in Spain in June?”. This change has only come about because AI has grown to the point where Google is able to understand this kind of conversational language, and is able to provide relevant search results based on these sorts of queries.

There are also practical search applications for the deep learning image analysis discussed above. One innovation that Google is working on is image searching as a shopping tool. For example, if you are interested in buying a specific belt, Google can then use its image search for images of people wearing similar belts and establishing the kinds of jeans that are typically worn with the belt – it can then make recommendations to you for jeans that you can potentially buy alongside the belt.

The challenges

Of course this is not to say that Google’s AI has reached its peak – nor that current AI technology is without its limitations. For example, it is likely now that the case that Google does not truly understand why its algorithm ranks websites in a particular way. This means that if a business feels that hard done by in Google’s search results, the company may no longer be able to provide a rational explanation for why a particular site ranks the way it does. Given the importance of Google to businesses across the UK, this might seem like a worrying prospect.

Additionally, it has been established that even with deep learning techniques, AI can be tricked or make mistakes. For example, Google’s image classification AI was tricked into believing a 3D printed turtle was a rifle.

There have even been concerns that AI is not able to think ethically in the way that humans are able to. For example, in late 2017 Facebook’s AI algorithm was discovered to have created anti-Semitic advert topics that users could target. This embarrassment forced Facebook to introduce human reviewers into something that had previously been a process that was completely automated by AI.

How will AI affect SEO?

Given the enormous variety of ways that AI is influencing search, it is clear that it will play a huge role in the SEO industry. In Part Two of this blog series on artificial intelligence and search engines we take an in-depth look into how AI is changing SEO.